Periodic transfer prediction method and device
By constructing time series data and feature maps, and using linear Transformer model and GNN graph neural network for training, the problem of poor prediction accuracy of periodic transfers in the existing technology is solved, and higher prediction accuracy is achieved.
Patent Information
- Application Number
- CN202411716947.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-03-18
AI Technical Summary
The prior art has poor accuracy in periodic transfer prediction and lacks effective identification and prediction methods.
By obtaining customer data, building time series data and feature maps, and using linear Transformer model and GNN graph neural network to train the data to generate periodic transfer prediction results.
The accuracy of periodic transfer prediction is improved, and the accuracy of model prediction is improved by combining time series data and interactive information.
Smart Images
Figure CN119647675A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of financial technology, and in particular to a method and device for predicting periodic transfers. Background Art
[0002] With the continuous development of my country's financial industry, completing the flow of funds through bank transfers has become a part of people's daily economic life. Some customers will transfer funds to an account at a fixed time every month. By identifying and predicting customers with periodic transfer needs, banks can guide customers to configure periodic transfer functions, understand customer needs and preferences, and improve customer stickiness. On the other hand, they can also better manage the flow of funds. At present, the identification of periodic transfer customers by banks is mainly completed by account managers through analyzing customer historical transfer records. The workload is large, the efficiency is low, and there is a lack of a complete set of rules. With the continuous advancement of artificial intelligence technology, the prediction algorithm for time series data such as transfer flow has made great progress. It has become practical to use machines to automatically identify and predict periodic transfer customers. Handing over repetitive and boring recognition and prediction tasks to computers can effectively improve the labeling efficiency and improve the consistency and quality of the identified data.
[0003] In the prior art, there are three main types of existing periodic transfer prediction methods. The first type is that bank account managers use existing experience based on the user's historical transfer behavior to manually determine whether the customer has periodic transfer behavior, which is inefficient and has poor accuracy. The second type is that data analysts write codes according to specific business rules and regard users who meet these rules as having periodic transfer behavior. The rules are fixed, with low flexibility and poor accuracy. The third type uses traditional machine learning algorithms, such as support vector machines, lightgbm, etc. The accuracy is improved compared to the second type of rule model, but the accuracy is still poor.
[0004] In summary, how to improve the accuracy of periodic transfer prediction is a problem that technical personnel in this field need to solve urgently. Summary of the invention
[0005] In view of this, the present application provides a periodic transfer prediction method and device, aiming to improve the accuracy of periodic transfer prediction.
[0006] In a first aspect, the present application provides a periodic transfer prediction method, comprising:
[0007] Obtaining customer data; the customer data includes customer transfer records and customer basic data;
[0008] Arrange customer transfer records in the order of timestamps to form time series data;
[0009] Taking customers as nodes, the nodes including the customer data, and taking the relationships and transaction patterns between the customers as connections between the nodes to construct a feature graph;
[0010] Inputting the time series data into a linear Transformer model to obtain a first output result;
[0011] Inputting the feature map into the GNN graph neural network to obtain a second data result;
[0012] A periodic transfer prediction result is generated according to the first output result and the second output result.
[0013] Optionally, after obtaining the customer data, the method further includes:
[0014] Identifying outliers in the customer data using a Z-score method;
[0015] Get the null value in the customer data;
[0016] Abnormal values and null values in the customer data are removed from the customer data.
[0017] Optionally, after arranging the customer transfer records in the order of timestamps to form time series data, the method further includes:
[0018] Performing feature extraction on the time series data to obtain normalized feature values of the time series data;
[0019] The step of inputting the time series data into a linear Transformer model to obtain a first output result includes:
[0020] The normalized feature values of the time series data are input into the linear Transformer model to obtain the first output result.
[0021] Optionally, before inputting the time series data into a linear Transformer model to obtain a first output result, the method further includes:
[0022] Build a linear Transformer model;
[0023] A self-attention mechanism is introduced into the linear Transformer model, so that the linear Transformer model establishes dependencies between sequence elements at different positions.
[0024] Optionally, generating a periodic transfer prediction result according to the first output result and the second output result includes:
[0025] The first output result and the second output result are input into a fully connected network to generate the periodic transfer prediction result.
[0026] Optionally, before inputting the first output result and the second output result into a fully connected network to generate the periodic transfer prediction result, the method further includes:
[0027] Construct a fully connected network.
[0028] In a second aspect, the present application provides a periodic transfer prediction device, comprising:
[0029] A first acquisition module is used to acquire customer data; the customer data includes customer transfer records and customer basic data;
[0030] The arrangement module is used to arrange the customer transfer records in the order of timestamps to form time series data;
[0031] A first construction module is used to use customers as nodes, the nodes include the customer data, and use the relationships and transaction patterns between the customers as connections between the nodes to construct a feature graph;
[0032] A first input module, used for inputting the time series data into a linear Transformer model to obtain a first output result;
[0033] A second input module is used to input the feature map into the GNN graph neural network to obtain a second data result;
[0034] A generating module is used to generate a periodic transfer prediction result according to the first output result and the second output result.
[0035] Optionally, the device further comprises:
[0036] an identification module, configured to identify outliers in the customer data using a Z-score method;
[0037] A second acquisition module, used to acquire a null value in the customer data;
[0038] The elimination module is used to eliminate abnormal values and null values in the customer data from the customer data.
[0039] Optionally, the device further comprises:
[0040] A feature extraction module, used to extract features from the time series data to obtain normalized feature values of the time series data;
[0041] The first input module is specifically used for:
[0042] The normalized feature values of the time series data are input into the linear Transformer model to obtain the first output result.
[0043] Optionally, the device further comprises:
[0044] The second building block is used to build a linear Transformer model;
[0045] The mechanism introduction module is used to introduce a self-attention mechanism into the linear Transformer model, so that the linear Transformer model establishes dependencies between sequence elements at different positions.
[0046] Optionally, the generating module includes:
[0047] A generating unit is used to input the first output result and the second output result into a fully connected network to generate the periodic transfer prediction result.
[0048] Optionally, the device further comprises:
[0049] A construction unit is used to construct a fully connected network.
[0050] In a third aspect, an embodiment of the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the periodic transfer prediction method described in any one of the implementation modes in the first aspect of the embodiment of the present application is implemented.
[0051] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes the periodic transfer prediction method described in any implementation method of the first aspect of the embodiment of the present application.
[0052] The present application provides a method for predicting periodic transfers. When executing the method, first obtain customer data, wherein the customer data includes customer transfer records and customer basic data, and then arrange the customer transfer records in the order of timestamps to form time series data. Next, take the customer as a node, which includes customer data, and take the relationship between customers and transaction patterns as the connection between the nodes to construct a feature graph, input the time series data into a linear Transformer model to obtain a first output result, input the feature graph into a GNN graph neural network to obtain a second data result, and finally generate a periodic transfer prediction result based on the first output result and the second output result. In this way, by combining the time series data and the interaction information between customers in the customer transaction process, and training the input features through a linear Transformer network and a GNN network, the accuracy of the model prediction is improved, thereby improving the accuracy of the periodic transfer prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] In order to more clearly illustrate the technical solutions in this embodiment or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0054] Figure 1 A flowchart of a periodic transfer prediction method provided in an embodiment of the present application;
[0055] Figure 2 A schematic diagram of the structure of a periodic transfer prediction device provided in an embodiment of the present application;
[0056] Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. The present application provides a method and device for predicting periodic transfers, which are used in the field of financial technology. The above is only an example and does not limit the application field of the method and device provided by the present application.
[0058] With the continuous development of my country's financial industry, completing the flow of funds through bank transfers has become a part of people's daily economic life. Some customers will transfer funds to an account at a fixed time every month. By identifying and predicting customers with periodic transfer needs, banks can guide customers to configure periodic transfer functions, understand customer needs and preferences, and improve customer stickiness. On the other hand, they can also better manage the flow of funds. At present, the identification of periodic transfer customers by banks is mainly completed by account managers through analyzing customer historical transfer records. The workload is large, the efficiency is low, and there is a lack of a complete set of rules. With the continuous advancement of artificial intelligence technology, the prediction algorithm for time series data such as transfer flow has made great progress. It has become practical to use machines to automatically identify and predict periodic transfer customers. Handing over repetitive and boring recognition and prediction tasks to computers can effectively improve the labeling efficiency and improve the consistency and quality of the identified data.
[0059] In the prior art, there are three main types of existing periodic transfer prediction methods. The first type is that bank account managers use existing experience based on the user's historical transfer behavior to manually determine whether the customer has periodic transfer behavior, which is inefficient and has poor accuracy. The second type is that data analysts write codes according to specific business rules and regard users who meet these rules as having periodic transfer behavior. The rules are fixed, with low flexibility and poor accuracy. The third type uses traditional machine learning algorithms, such as support vector machines, lightgbm, etc. The accuracy is improved compared to the second type of rule model, but the accuracy is still poor.
[0060] After research, the inventor proposed the technical solution of this application. First, customer data is obtained, including customer transfer records and customer basic data. Then, the customer transfer records are arranged in the order of timestamps to form time series data. Next, the customer is used as a node, which includes customer data, and the relationship and transaction mode between customers are used as the connection between the nodes to construct a feature graph. The time series data is input into the linear Transformer model to obtain the first output result, and the feature graph is input into the GNN graph neural network to obtain the second data result. Finally, according to the first output result and the second output result, a periodic transfer prediction result is generated. In this way, by combining the time series data and the interaction information between customers in the customer transaction process, and training the input features through the linear Transformer network and the GNN network, the accuracy of the model prediction is improved, thereby improving the accuracy of the periodic transfer prediction.
[0061] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application. It should be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings. In the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.
[0062] See also Figure 1 , Figure 1 A flowchart of a periodic transfer prediction method provided in an embodiment of the present application includes:
[0063] S101: Acquire customer data.
[0064] Obtain customer data, where customer data includes customer transfer records and customer basic data. Customer transfer records include information such as the timestamp of each transfer, transfer amount, and transfer object. Customer basic data includes information such as user age and gender.
[0065] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.
[0066] In some implementations, the Z-score method may be used to identify outliers in customer data. The specific calculation method is as follows:
[0067]
[0068] Among them, X is the value of the data point, μ is the mean of the data, and σ is the standard deviation of the data. When the absolute value of the Z score is greater than 3, the data is confirmed as an outlier, and the outliers and null values in the customer data are eliminated.
[0069] S102: Arrange the customer transfer records in the order of timestamps to form time series data.
[0070] Arrange the transfer records of each customer in the order of timestamps to form time series data. In some implementations, features can be extracted from the time series data, including the average and maximum transfer amounts in the past 1, 2, and 3 months, the average and maximum transfer frequencies in the past 1, 2, and 3 months, etc. Each feature value is scaled to between [0, 1] using the decimal scaling method, as follows:
[0071]
[0072] Among them, X max is the maximum value of the feature, X norm is the normalized eigenvalue.
[0073] S103: Take customers as nodes, the nodes include customer data, and take the relationships and transaction patterns between customers as connections between nodes to construct a feature graph.
[0074] By treating each customer as a node, each node contains the customer's basic data, mainly customer ID, age, gender and other basic user information. The relationship and transaction pattern between customers are used as the connection between nodes. For example, if customer a transfers money to customer b twice, there is a connection between a and b, and the connection weight is 2. The specific calculation method is:
[0075]
[0076] S104: Input the time series data into the linear Transformer model to obtain a first output result.
[0077] Before executing step S104, it is necessary to construct a linear Transformer model. Specifically, while maintaining the core idea of Transformer, the model is allowed to establish dependencies between sequence elements at different positions by introducing a self-attention mechanism. The input of the self-attention mechanism includes query, key, and value, where Query is the input information of the current sequence position, Key is the input information of other sequence positions, and Value is the output information of other sequence positions. Calculate the attention distribution α ij The method is as follows:
[0078]
[0079] Among them, Q i and K j They are the vector representations of Query and Key respectively, and d is the dimension of Query and Key.
[0080] Attention weight β ij The calculation method is as follows:
[0081] β ij =softmax(α ij );
[0082] Finally, calculate the aggregate value γ i :
[0083] γi =∑ j β ij V j ;
[0084] Among them, V j It is the vector representation of Value.
[0085] In addition to the self-attention mechanism, a feedforward neural network is also introduced into the network. This network mainly implements nonlinear transformation and maps the output results of the multi-head attention layer to a higher-dimensional space. The calculation method is:
[0086] FFN=max(0,xW1,b1)W2+B2;
[0087] Among them, x is the output of the previous layer, and W, b, B are offsets.
[0088] Before the self-attention mechanism and feedforward neural network, the model also introduces residual connections and normalization layers to alleviate the gradient vanishing and gradient exploding problems that the model may encounter and stabilize the model training process.
[0089] S105: Input the feature map into the GNN graph neural network to obtain a second data result.
[0090] The input to the GNN network is a feature map that has been constructed. Through the three-layer GNN, the model can capture the local and global relationships between nodes, helping the model understand the interactions and dependencies between customers.
[0091] S106: Generate a periodic transfer prediction result according to the first output result and the second output result.
[0092] The first output result and the second output result are input into the fully connected network to generate a periodic transfer prediction result.
[0093] In an embodiment of the present application, customer data is first obtained, wherein the customer data includes customer transfer records and customer basic data, and then the customer transfer records are arranged in the order of timestamps to form time series data. Next, the customer is used as a node, which includes customer data, and the relationship between customers and transaction patterns are used as connections between the nodes to construct a feature graph, and the time series data is input into a linear Transformer model to obtain a first output result, and the feature graph is input into a GNN graph neural network to obtain a second data result. Finally, a periodic transfer prediction result is generated based on the first output result and the second output result. In this way, by combining the time series data in the customer transaction process and the interaction information between customers, and training the input features through a linear Transformer network and a GNN network, the accuracy of the model prediction is improved, thereby improving the accuracy of the periodic transfer prediction.
[0094] The above are some specific implementations of the periodic transfer prediction method provided by the embodiment of the present application. Based on this, the present application also provides a corresponding device. The device provided by the embodiment of the present application will be introduced from the perspective of functional modularization.
[0095] See also Figure 2 , Figure 2 A schematic diagram of the structure of a periodic transfer prediction device provided in an embodiment of the present application, the periodic transfer prediction device 200 includes:
[0096] A first acquisition module 210 is used to acquire customer data; the customer data includes customer transfer records and customer basic data;
[0097] Arrangement module 220, used to arrange customer transfer records in the order of timestamps to form time series data;
[0098] A first construction module 230 is used to use customers as nodes, the nodes include the customer data, and use the relationships and transaction patterns between the customers as connections between the nodes to construct a feature graph;
[0099] A first input module 240, used to input the time series data into a linear Transformer model to obtain a first output result;
[0100] A second input module 250 is used to input the feature map into the GNN graph neural network to obtain a second data result;
[0101] The generating module 260 is used to generate a periodic transfer prediction result according to the first output result and the second output result.
[0102] Optionally, the device 200 further includes:
[0103] an identification module, configured to identify outliers in the customer data using a Z-score method;
[0104] A second acquisition module, used to acquire a null value in the customer data;
[0105] The elimination module is used to eliminate abnormal values and null values in the customer data from the customer data.
[0106] Optionally, the device 200 further includes:
[0107] A feature extraction module, used to extract features from the time series data to obtain normalized feature values of the time series data;
[0108] The first input module 240 is specifically used for:
[0109] The normalized feature values of the time series data are input into the linear Transformer model to obtain the first output result.
[0110] Optionally, the device 200 further includes:
[0111] The second building block is used to build a linear Transformer model;
[0112] The mechanism introduction module is used to introduce a self-attention mechanism into the linear Transformer model, so that the linear Transformer model establishes dependencies between sequence elements at different positions.
[0113] Optionally, the generating module 260 includes:
[0114] A generating unit is used to input the first output result and the second output result into a fully connected network to generate the periodic transfer prediction result.
[0115] Optionally, the device 200 further includes:
[0116] A construction unit is used to construct a fully connected network.
[0117] It can be understood that in the specific implementation of the present application, related data such as customer transfer records and customer basic data are involved. When the above embodiments of the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.
[0118] The embodiments of the present application also provide corresponding devices and computer storage media for implementing the solutions provided by the embodiments of the present application.
[0119] like Figure 3 As shown, the computer device 01 is in the form of a general computing device. The components of the computer device 01 may include but are not limited to: one or more processors or processing units 03, a system memory 08, and a bus 04 connecting different system components (including the system memory 08 and the processing unit 03).
[0120] Bus 04 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnect (PCI) bus.
[0121] The computer device 01 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 01, including volatile and non-volatile media, removable and non-removable media.
[0122] The system memory 08 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 09 and / or cache memory 10. The computer device 01 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 11 may be used to read and write non-removable, non-volatile magnetic media ( Figure 3 not shown, usually called a "hard drive"). Although Figure 3 Not shown in the figure, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 04 via one or more data medium interfaces. The memory 08 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.
[0123] A program / utility 12 having a set (at least one) of program modules 13 may be stored, for example, in the memory 08, such program modules 13 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination thereof may include an implementation of a network environment. The program modules 13 generally perform the functions and / or methods of the embodiments described herein.
[0124] The computer device 01 may also communicate with one or more external devices 02 (e.g., keyboard, pointing device, display 07, etc.), one or more devices that enable a user to interact with the computer device 01, and / or any device that enables the computer device 01 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via input / output (I / O) interface 06. Furthermore, the computer device 01 may also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN) and / or public network, such as the Internet) via network adapter 05. Figure 3 As shown, the network adapter 05 communicates with other modules of the computer device 01 via the bus 04. It should be understood that although Figure 3Not shown, other hardware and / or software modules may be used in conjunction with computer device 01, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0125] The processor unit 03 executes various functional applications and data processing by running the programs stored in the system memory 08, for example, implementing a periodic transfer prediction method provided in an embodiment of the present application.
[0126] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0127] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a read-only memory (ROM) / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment of the present application or some parts of the embodiments.
[0128] Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.
[0129] The above description is merely an exemplary embodiment of the present application and is not intended to limit the protection scope of the present application.
Claims
1. A periodic transfer prediction method, characterized in that: include: Obtaining customer data; the customer data includes customer transfer records and customer basic data; Arrange customer transfer records in the order of timestamps to form time series data; Taking customers as nodes, the nodes including the customer data, and taking the relationships and transaction patterns between the customers as connections between the nodes to construct a feature graph; Inputting the time series data into a linear Transformer model to obtain a first output result; Inputting the feature map into the GNN graph neural network to obtain a second data result; A periodic transfer prediction result is generated according to the first output result and the second output result.
2. The method according to claim 1, characterized in that After obtaining the customer data, the method further includes: Identifying outliers in the customer data using a Z-score method; Get the null value in the customer data; Abnormal values and null values in the customer data are removed from the customer data.
3. The method according to claim 1, characterized in that After arranging the customer transfer records in the order of timestamps to form time series data, the method further includes: Performing feature extraction on the time series data to obtain normalized feature values of the time series data; The step of inputting the time series data into a linear Transformer model to obtain a first output result includes: The normalized feature values of the time series data are input into the linear Transformer model to obtain the first output result.
4. The method according to claim 1, characterized in that: Before inputting the time series data into the linear Transformer model to obtain the first output result, the method further includes: Build a linear Transformer model; A self-attention mechanism is introduced into the linear Transformer model, so that the linear Transformer model establishes dependencies between sequence elements at different positions.
5. The method according to claim 1, characterized in that Generating a periodic transfer prediction result according to the first output result and the second output result includes: The first output result and the second output result are input into a fully connected network to generate the periodic transfer prediction result.
6. The method according to claim 5, characterized in that Before inputting the first output result and the second output result into a fully connected network to generate the periodic transfer prediction result, the method further includes: Construct a fully connected network.
7. A periodic transfer prediction device, characterized in that: include: A first acquisition module is used to acquire customer data; the customer data includes customer transfer records and customer basic data; The arrangement module is used to arrange the customer transfer records in the order of timestamps to form time series data; A first construction module is used to use customers as nodes, the nodes include the customer data, and use the relationships and transaction patterns between the customers as connections between the nodes to construct a feature graph; A first input module, used for inputting the time series data into a linear Transformer model to obtain a first output result; A second input module is used to input the feature map into the GNN graph neural network to obtain a second data result; A generating module is used to generate a periodic transfer prediction result according to the first output result and the second output result.
8. The method according to claim 7, characterized in that The device also includes: an identification module, for identifying outliers in the customer data using a Z-score method; A second acquisition module, used to acquire a null value in the customer data; The elimination module is used to eliminate abnormal values and null values in the customer data from the customer data.
9. A computer device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the periodic transfer prediction method according to any one of claims 1 to 6 is implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the periodic transfer prediction method according to any one of claims 1 to 6.